The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill takes user-provided descriptions and interpolates them into prompts for AI models (e.g., bytedance/seedream-4-5). This creates a surface where instructions embedded in the input data could potentially influence the model's output or the agent's behavior.
Ingestion points: User-defined product attributes, categories, and descriptive keywords integrated into the belt CLI commands in SKILL.md.
Boundary markers: Absent; the skill uses direct string interpolation into JSON payloads for the CLI tools without specific delimiters or "ignore instructions" headers.
Capability inventory: The skill uses the Bash(belt *) tool to execute shell commands and interact with remote AI inference services.
Sanitization: No validation or escaping of user-provided content is implemented in the provided prompt templates.
[EXTERNAL_DOWNLOADS]: The skill requires and recommends the installation of the belt-sh/cli and references external documentation and installation scripts located on GitHub.
Evidence: The skill points to https://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md for installation instructions and uses npx skills add belt-sh/cli to install the primary tool dependency.

product-photography